Author: Edmund L. Andrews
Publisher: Stanford University
Publication Year: 2021
Summary: The following article discusses how artificial intelligence (AI) models used to determine credit worthiness for lending are less accurate for lower-income and minority borrowers, in part due to less information on their credit reports. This leads to fewer loans to these borrowers and perpetuates the problem that there is little information about borrowing history upon which to evaluate their risk. Another problem with having minimal positive information to use for evaluation is that any negative information will have an outsized impact on the credit score. “One possible [solution] would be for financial companies to run experiments in which they approve loans to people with relatively low credit scores,” thus building a dataset for better models in the future.